Tianle LI

Tianle LI

First Year Master Student

University of Waterloo

Biography

My research interests include Natural Language Processing and Image Editing conditioned on text. Currently, I am a thesis-based CS master student in University of Waterloo supervised by Prof. Wenhu Chen. Collaborating with GPT-series and diffusion-based models, we are exploring the potentials of the popular large language and vision models in various tasks, designing efficient algorithm on existed or novel problem setting. Ultimately, I hope we can realize more controllable generation, grounding on truth or human intention.

Interests

  • Natural Language Processing (Biasness, Understanding, Robustness)
  • Computer Vision (Image Editing/Generation Conditioned on Text)
  • Scoial Network Mining

Education

  • Computer Science - MMath, 2022

    University of Waterloo

  • BSc in Data Science and Technology (DSCT), 2017-2021

    The Hong Kong University of Science and Technology

Skills

Python

Tensorflow, Pytorch, Pandas, Spacy, etc.

Other Programming Language

C++, JAVA, SQL, MATLAB, etc.

Cloud Platform

AWS, GCP

Experience

 
 
 
 
 

Research Intern

ALIBABA DAMO ACADAMY

Mar 2022 – Aug 2022 Hangzhou, China

Supervised by Jing Wu and Boxing Chen

  • Design and implement a simultaneous speech translation model POST, which mitigetes the reordering issue when translating the language pairs in real-time by predicting the position of each word.
 
 
 
 
 

Research Intern

Microsoft Research Lab - Asia

Feb 2021 – Sep 2021 Beijing, China

Supervised by senior researcher Mengyu Zhou

  • Design and implement the advanced algorithm for key phrases extraction. (This work has been deployed in Forms, Teams as Word cloud insight to provide a visualization of responses for text questions. It recieved very positive feedbacks from the customers!)
  • Learning Analysis Semantics over Tabular Data via Conditional Formatting as Proxy. In this research project, we proposed a sophiscated method to encode the table context with good comprehension to numerical tabular data, which can select condition type, formatting type and critical values to automatically recommend conditional formatting visualization for human beings.
 
 
 
 
 

Final Year Project: Fake News Detection on Social Networks

HKUST

Sep 2020 – Sep 2021 Hong Kong
Under the supervision from Professor Raymond, Wong, we propose a novel Transformer-based model: HetTransformer to solve the fake news detection problem on social networks, which utilizes the structure-aware Transformer and temperal embedding to capture the news propagation patterns in social media. Experiments on three real-world datasets demonstrate that our model is able to outperform the state-of-the-art baselines in fake news detection.
 
 
 
 
 

Junior Research Assistant

HKUST

Jun 2020 – Aug 2020 Hong Kong
We aim at generating adversarial examples in text to attack pretrained BERT model in black box setting under budget constraint (merely query much less number of times towards target model with the same level of success rate and perturbation rate). We employ all the intermediate failure and successful queries to learn words salience rank globally and locally.
 
 
 
 
 

Summer Research Intern

Hong Kong Applied Science and Technology Research Institute (ASTRI)

Jun 2019 – Aug 2019 Hong Kong
Under the supervision from Abel Ze, Yang, we utilized cross-chain blockchain technology and recurrent neural net network to build a cryptocurrency exchange rate prediction system, which obtained recognition from our department head James Zhibin, Lei.
 
 
 
 
 

Undergraduate Research Project

HKUST

Feb 2019 – Aug 2019 Hong Kong
Under the supervision from Prof. Jian-Feng, Cai, I exploited accelerated alternating projections for robust PCA and low-rank Hankel matrix completion to efficiently decompose a Hankel matrix into a low-rank Hankel matrix and a noisy sparse matrix with MATLAB. After the design of the algorithm, I conducted experiments with real data from stock market which leads to 2% improvement on price prediction accuracy.

Accomplish­ments

Participant

I worked in a team and implemented a person portrait matting algorithm based on UNet-structure network. We further built a style transfer system on matting person portrait images with utilization of ResNet and CycleGAN to transfer images from realistic style to Simpsons style on GCP.

Top 10

I worked in a group to Scrape reviews from HK.trip.com and classify positive and negative keywords for distinct targeted customers through sentiment analysis and visualize it with word clouds. With the collection of the datasets and basic analysis, we exploited deep-wide learning model to recommend desired hotels for online customers.

Winner

I led a team to build a management system for Premiere Performances, which is a NGO, to implemente search function, nodes graph visualization, engagement and likeness of clients in the database. And we won the bid from Premiere Performances with recognition of managers from J.P. Morgan.

Contact

  • tliax@connect.ust.hk
  • (852) 95653901
  • HKUST Jockey Club Hall, 3 Tong Yin Ln, Tseung Kwan O, Hong Kong,